Understanding mammalian biology and disease through tissue-specific protein-protein interaction networks

Biological functions are mediated by the dynamic organization of DNA, RNA, proteins, and other biomolecules in complex networks of interactions. Efforts to chart the network of biologically relevant macromolecular interactions—the “interactome”—therefore occupy a central position in the endeavour to understand the biochemical basis of human physiology, and its perturbation in disease. However, existing interactome maps are incomplete, even for well-studied organisms. Moreover, the dynamics of the interactome in response to cellular stimuli and across normal physiological contexts remain incompletely understood. This thesis considers the application of a quantitative proteomic approach, protein correlation profiling (PCP), to map the interactome in its native physiological context. I explore computational methods for the analysis of PCP data, and describe their application to infer a dataset of protein-protein interactions from seven mouse tissues. In Chapter 2, I studied the dominant paradigm used to analyze PCP data, which entails the use of supervised machine-learning methods to infer interaction networks from these complex datasets. I found that one widely used strategy needlessly biases network inference towards highly studied proteins and away from novel interactions between functionally un-connected proteins. In Chapter 3, I applied the methods studied in Chapter 2 to a newly collected, in vivo PCP dataset. I used the same machine-learning approach to infer tissue-specific protein-protein interaction networks for seven mouse tissues. I then analyzed these tissue interactome networks to uncover insights about protein function, network evolution, and human disease. Collectively, the work described in this thesis provides a framework to understand the rewiring of the protein-protein interaction network across physiological conditions using PCP.Science, Faculty ofGraduat